Presentation | 2003/12/1 A Method to Estimate the Learning Coefficients of Singular Learning Machines by Decomposition of Kullback Information Kenji NAGATA, Sumio WATANABE, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | A lot of learning machines such as neural networks, normal mixtures, Bayesian networks, and hidden Markov models are singular statistical models. Their Fisher information matrices are not positive definite, hence the conventional statistical asymptotic theory does not hold. In this paper, we propose a new method to calculate the learning coefficients by decomposing the Kullback information. The effectiveness of the proposed method is shown by experimetal results. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Singular Learning Machines / Kullback Information / Stochastic Complexity |
Paper # | NC2003-104 |
Date of Issue |
Conference Information | |
Committee | NC |
---|---|
Conference Date | 2003/12/1(1days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | |
Chair | |
Vice Chair | |
Secretary | |
Assistant |
Paper Information | |
Registration To | Neurocomputing (NC) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | A Method to Estimate the Learning Coefficients of Singular Learning Machines by Decomposition of Kullback Information |
Sub Title (in English) | |
Keyword(1) | Singular Learning Machines |
Keyword(2) | Kullback Information |
Keyword(3) | Stochastic Complexity |
1st Author's Name | Kenji NAGATA |
1st Author's Affiliation | Department of Computer Science Tokyo Institute of Technology() |
2nd Author's Name | Sumio WATANABE |
2nd Author's Affiliation | PI Lab., Tokyo Institute of Technology |
Date | 2003/12/1 |
Paper # | NC2003-104 |
Volume (vol) | vol.103 |
Number (no) | 490 |
Page | pp.pp.- |
#Pages | 6 |
Date of Issue |